Internet has become essential to human nowadays. People can communicate easily using internet from almost any places in the world. The internet has rapidly grows over 10 years. Statistic shows that estimated that there were 1.8 billion internet users compared to 360 million users in year 2000. The percentage of internet users has growth over 399.3% over this ten years.As the internet keep expanding and growing rapidly in terms of size and traffic load, the number of routers in a routing domains is also become larger. This situation eventually leads to frequent topological changes because of link failures, recoveries and changes.
1.1.1 Router and Routing Protocol
Router can be defined as a device that determines the next network point to which a packet should be forwarded towards its destination . The routing protocol use in the routers specifies how the routers communicate with each others. They are two major classes of routing protocol which are the interior gateway routing protocol
and the exterior gateway routing protocol. In this research we will be focusing on a particular interior gateway routing protocol called Open Shortest Path First (OSPF).
The interior gateway routing protocol operates within an autonomous system and can be divide into two categories which is distance vector routing protocol and link-state routing protocol. Distance vector routing protocol informs their neighbors about topology changes periodically.In the distance vector routing protocols, routers discover the best path to the destination from their each neighbors. Unlike the distance vector routing protocol, the link-state routing protocol require routers to inform all the nodes in an area about topology changes.
Basically node maintains a map of connectivity each for the area it resides. When topology change occurs, the affected nodes will have to recompute the best path to destinations. The collection of the best path will form the nodeâ€™s routing table. Once the entire nodeâ€™s routing table has been updated, convergence is complete.
1.1.2 Convergence Process and Topology Changes
Convergence is the process of routers agreeing on optimal routers for forwarding packets and thereby completing the updating of their routing tables.It is important for routing protocol to have a quick convergence nowadays because of real time applications for example video-conferencing and VoIP (Voice over Internet Protocol). The link-state routing protocol provide greater flexibility and sophistication compared to the distance vector routing protocol. In term of speed of convergence, distance vector routing protocol can converge slowly and have routing loops while converging and also suffers from count to infinity problems .
This research will be focusing on the one of the link-state routing protocol which is Open Shortest Path First (OSPF). OSPF has become famous routing protocol and most of Internet Service Providers (ISP) use OSPF as their interior gateway routing protocol. OSPF used Link State Advertisement (LSA) to describe the local state of a router or network and for a router, LSA also describing the current state of the routerâ€™s interfaces and adjacencies . At the event of topology changes, the nodes will sent the LSA to the all nodes within autonomous system. The nodes than will calculate the new routing table.
There were two routing table calculation scheduling schemes in OSPF which is hold time based scheme and LSA Correlation. The Hold-time based scheme uses value of delay parameters that configured by the network administrator. For LSA Correlation, it assumes that new LSA does not trigger new routing table calculation but assume that it was symptom of a topology changes. So, the routers should a router should correlate the information in individual new LSA to identify the topology change itself and then perform a routing table calculation .
This scheme has just been proposed and the topology change identification is not straight-forward. The Hold-time based and LSA Correlation has not used computational intelligence techniques. In this research, we propose the use of computational intelligence in the new scheduling of routing table for optimizing the frequency of routing table calculation
1.1.3 Computational Intelligence
Fuzzy Logic and Artificial Neural Network are examples of computational intelligence techniques. For this research, we will be using artificial neural network to optimize the scheduling of routing table calculation schemes. Artificial Neural Network (ANN) is a computational model that tries to simulate biological process of the human brain. Basically ANN consist 3 types of layer which is input layer, hidden layer and output layer. The function of ANN is to process information, and it suitable to process the scheduling of routing table information to determine the suitable frequency of routing table calculations.
The hold-time based scheme is divided into two types which is fixed hold-time and exponential back-off hold time schemes. In this scheme, routers not in the convergence process are in initial state. When the routers that in initial state received LSA, the routers will change state to the SPF (Shortest Path First) state where the routing table calculation were initiated immediately and hold timer is started. It assumed that the time needed for a routing table calculation is less than the hold time. In the SPF state, the routers will wait for the hold time to expire or the arrival of the new LSA. When the hold time is expiring, it will cause the routers to return to the initial state. However the router that in SPF state received new LSA, that router will change state to the SPF hold state. In the SPF hold state, the router has one or more pending LSA and is waiting for the hold time to expire so that the router can return to the SPF state and perform a routing table calculation. This is when all the received LSA will be process while the hold timer is running.
In the exponential back-off hold time scheme, the transition state from the SPF state to the SPF Hold state causes the hold time to double in value up to maximum. The scheme starts with a small value for hold time. If the LSA received frequently, the hold time will quickly reach its maximum value therefore limiting the frequency of routing table calculation. However, it is possible that no LSA is received during hold time duration and the hold time will be reset to its small initial value.
The both schemes depend on the values of several delay parameters. The main problems of the both method is to determine the values for the parameter that will result in fast convergence time for all possible topology change scenarios that happen to a network.
The other schemes for routing table calculation scheduling are LSA Correlation. A router randomly receives new LSA at one or more of its interface because of topology changes in network. Each new LSA will be update to Link-State Database (LSDB) that stores the map of the routers is OSPF by the routers. The content of the new LSA will be examined, correlated and compared to the previous LSA when necessary to decide whether a topology changes can be determine. If a topology change has been identified and determine, then the new routing table calculation needs to be scheduled. In LSA Correlation, many scenarios need to be handled and extra memory space is needed for the new and older LSA for the topology identification process.
This research will propose scheduling schemes that is less complicated than LSA Correlation but still offered minimum frequency of routing table calculation and fast convergence time. To the best of our knowledge, no scheduling schemes for routing table calculation that used computational intelligence so far. This research will explore the possibility of using computational intelligence to solve the research problems.
To support the research problems that been discussed in section 1.2, this research is conducted to provide the answers to the following questions:
How can we model the scheduling of routing table calculation using artificial neural network.
How can we use artificial neural network model to minimize the frequency of routing table calculations.
Does artificial neural network provides a better performance compared to existing scheduling mechanism.
1.4 Dissertation Aim
This research aimed to speeding up OSPF convergence time by minimizing the frequency of routing table calculation. In this research, we propose a new scheduling scheme that use Artificial Neural Network model to optimize the scheduling of routing table calculation.
1.5 Dissertation Objectives
The objectives for this research are stated as below:
To develop Artificial Neural Network model for the proposed schemes.
To incorporated Artificial Neural Network Model in a proposed schemes.
To perform evaluation report between the proposed schemes.
1.6 Scope of the Study
To achieve research objectives, it is important to determine research scope and the limits to the research. The scope of this research is:
This research will focus on speeding up OSPF convergence time by optimizing the scheduling of routing table calculation.
To optimizing the scheduling of routing table calculation, the proposed scheme will use Artificial Neural Network model.
1.7 Significance of the Study
This project will study on scheduling of routing table calculations using artificial neural network. This was the first time the artificial intelligence implemented to schedule the routing table. If this research become success, it can be implemented to the commercial routers and it will benefits the users of the internet. When the time taken by OSPF to convergence is decreases, the network performance will increase thus satisfies the demand of fast failure recovery applications such as Voice over Internet Protocol and live video streaming.
1.8 Dissertation Organization
There are 6 chapters in this dissertation report. The first chapter provides the introduction of the study, problem background and statement, aim, objectives, scope of the studies and the significance of this research. Chapter 2 will gives the literature review about topics related to the research. Methodology use in the research will be discussed in the chapter 3. Chapter 4 will explained about the proposed method of the research while chapter 5 will present research results. Chapter 6 is where the conclusion and suggestion for future works is stated.